perf(ws): reduce needless allocations

[JohnTitor]

PR Type

INSERT_PR_TYPE

PR Checklist

• Tests for the changes have been added / updated.
• Documentation comments have been added / updated.
• A changelog entry has been made for the appropriate packages.
• Format code with the nightly rustfmt (cargo +nightly fmt).

Overview

• reuse total_len instead of caluculating
• prefer BytesMut::split over mem::take to avoid reallocations
• skip calling extend_from_slice to avoid unnecessary copy

[asonix]

Using mem::take here was to prevent the BytesMut from growing unbounded with large payloads and never shrinking again.

See: JohnTitor@8195484
And: actix/actix-web#3367

[JohnTitor]

@asonix Ah, that makes sense, thank you for pointing out! I'll partially revert it and add a note later.